Spreadsheets are remarkably flexible things. They are probably the single most common vector for "citizen programming" in the world at large, and they can easily do simple databaselike tasks, as well as their intended calculation roles. That said, I find it surprising that Excel looks so very much like its ancient ancestor, Visicalc. Occasionally slightly cooler graphing packages or actual new ways of data manipulation show up, but it's been rare over the years.
In 2004 or so I sat through a presentation by a gentleman from Microsoft, explaining that the next step in the evolution of the Office suite was collaboration tools. That's very nice, but there were so many times I would have torn off my own arm to get more than 256 columns and more than 65536 rows into the damn thing.
Instead, having settled on the exact form of the thing, we reject new ways to do the same tasks. Simple features like being able to rename columns and use those columns in formulas, an innovation first shown in Lotus Improv, was rejected and dissapeared long ago: why? Examine Christopher Browne's history for a taste of what could have been.
So when I saw Google's recent entry into the web-based productivity suite I was a bit surprised and a bit saddened, and not at all by what it does but instead by what it does not do.